WESTCARB Regional Partnership Subsurface Flow Modeling at King Island Christine Doughty, CADoughty@lbl.gov Curtis Oldenburg, CMOldenburg@lbl.gov Staff Scientists Lawrence Berkeley National Laboratory WESTCARB Annual Business Meeting Bakersfield, CA October 15 17, 2012 1
Outline Objectives Model Development Simulation Results Assessment of Data Use/Information Flow Future Work 2
Objectives King Island Project as a Whole Assess the suitability of the Southern Sacramento Basin for CO2 sequestration Apply a variety of existing site characterization methods Develop new site characterization methods Numerical Modeling Predict movement and trapping of injected CO2 in the subsurface Assess storage capacity Estimate risk of leakage and pressure increase Evaluate value of various site characterization methods for providing input to model 3
Location Map Central Valley Drill site Southwestern Sacramento Valley provides closest potential storage site for San Francisco Bay Area refineries 4
Model Development Target Storage Formations Structure Lateral Extent and Boundary Conditions Material Properties Initial Conditions Representation of Well 5
Target Storage Formations Note: gorge fill has shale-like properties; effectively creates undulations in sandbody caprock 6
Model Structure Ignore gorges Approximate dip as uniform dip: 1.6 o (up is to the ENE) Model is a tilted plane From Downey and Clinkenbeard, 2011 7
Lateral Extent and Boundary Conditions Axes aligned with dip direction West and South boundaries approximately aligned with faults East boundary open as Mokelumne may abut another permeable formation Closed Closed y Constant Pressure Closed x 8
Conceptualization of Entire Depth Interval (prior to drilling Citizen Green Well) Total thickness 1280 m Rough estimates and literature values Sands Porosity 0.25-0.35 Permeability 50 500 md Anisotropic: k v /k h 0.01 to 0.1 Shales Porosity 0.10 Permeability 10 md Anisotropic: k v /k h 0.1 Properties are effective values that account for sub-grid-scale heterogeneity Vertical crosssection of 3D model (aligned along dip direction) 9
Focus on Mokelumne River Sandstone (incorporating data from Citizen Green Well) Total thickness 483 m Well log and sidewall core comparison (J.B. Ajo-Franklin) indicates: NMR Total Porosity appears to best match helium porosimetry data from the sidewall samples NMR permeability estimates appear to be relatively accurate for the formations encountered at the Citizen Green well 10
Well Logs and Model Discretization Model layering from well-log permeability profile Sharp boundary to overlying Capay Shale - no-flow boundary a good approximation Layer properties Porosity and horizontal permeability - arithmetic mean Vertical permeability - harmonic mean Big variation of permeability > three orders of magnitude Model does not capture all detail, but hopes to represent key features High permeability in upper half Downward fining in lower half Low-permeability baffles over whole thickness greatly decrease effective vertical permeability 3D model, 37,620 grid blocks 483 m thick, 19 layers Lateral grid resolution near well 50 by 50 m, coarsens outward Lateral extent of model 42 by 60 km 11
TOUGH2 Numerical Simulator Fully coupled multiphase fluid flow Equation of state includes H 2 O, CO 2, NaCl Isothermal simulations Injected CO 2 forms a gas-like supercritical phase and dissolves in brine Variety of capillary pressure and relative permeability functions available Can be fit to literature or laboratory data Here use generic characteristic curves 12
Initial Conditions Single-phase liquid brine (salinity 50,000 ppm) Hydrostatic pressure profile (86-250 bars) Geothermal temperature gradient (50-80 o C) 13
Representation of CO 2 Injection Diagonal well represented by stairsteps in rectangular grid Assume well perforated over the lower half of the Mokelumne River Injection partitioned among grid block representing well according to permeability- thickness product Does not account for different pressure gradients in well (CO 2 ) and formation (brine) Inject 1 MT CO 2 per year for four years Over-estimates injection at greater depth 14
Simulation Results scco 2 saturation Strong preferential flow in highpermeability layers Strong buoyancy flow within highpermeability layers Slight up-dip migration 15
Simulation Results Pressure Change Distributions Pressure increase moderate for highpermeability formation Extent of pressure change much greater than extent of CO 2 16
CO 2 Mass Balance for Entire Model Injected CO 2 forms a supercritical phase and dissolves into the aqueous phase Dissolved fraction is consistently about 25% 17
Data Use/Information Flow Model 1 Regional geology Inject into four sand layers Model 2 Anecdotal information from lack of core recovery from Citizen Green well: high permeability in top Moke Inject into lower half Moke Strong buoyancy flow Model 3 Well logs from Citizen Green well Sidewall core from Citizen Green well Inject into lower half Moke Layering restricts buoyancy flow 18
Future Work Incorporation of additional data Include lateral heterogeneity Well logs available from 4 nearby wells Use Citizen Green data to calibrate Incorporate realistic characteristic curves Micro CT and analytical solution CO 2 /brine flood sequence in Domengine core - seismic response shows strong hysteresis Simulations Injection period: growth and movement of plume (pressure-driven and buoyancy flow) Post-injection: plume evolution after injection ends (buoyancy flow, trapping mechanisms) 19